---
title: 'In the News: September 11, 2026'
description: 'Sakana AI prices Fugu Max at $2 per million input tokens and $6 per million output tokens, 40 to 60 percent below three named coding models.'
canonical_url: 'https://darkfactory.dev/news/2026-09-11-extra-2'
markdown_url: 'https://darkfactory.dev/news/2026-09-11-extra-2.md'
collection: news
date_published: '2026-09-11T10:12:07-04:00'
date_modified: '2026-09-11T10:12:07-04:00'
---

# In the News: September 11, 2026


Sakana AI prices Fugu Max at $2 per million input tokens and $6 per million output tokens. The company says its output price is 40 to 60 percent below Sonnet 5, GPT 5.6 Terra, and Kimi K3.

## 1. Sakana AI's Fugu Max undercuts flagship models on agentic coding price

**[Introducing Fugu Max and Fugu Ultra v2: Orchestrating the Pareto Frontier](https://sakana.ai/fugu-max-release/)** · Sakana AI · sakana.ai, September 11, 2026

Sakana AI released Fugu Max and Fugu Ultra v2, two configurations of its multi-agent orchestration system. The system routes tasks across a pool of open and specialized models instead of relying on one frontier model.

Fugu Max costs $2 per million input tokens and $6 per million output tokens. Sakana says its output price is 40 to 60 percent below Sonnet 5, GPT 5.6 Terra, and Kimi K3. The company also reports the best overall score on six benchmarks, including Terminal Bench 2.1 and its own internal SWEFish coding benchmark.

Fugu Ultra v2 targets complex multi-step reasoning and full-stack software development. It scores 74.3 on the DeepSWE software-engineering benchmark, which Sakana says beats models costing three to five times as much per token.

Both models use Sakana's existing OpenAI-compatible endpoint. The company says current users can switch with "a single-line parameter change."

**Why it matters:** Teams routing coding-agent traffic by price now have an option priced at $2 per million input tokens and $6 per million output tokens, measured against three named models. Sakana reports the benchmark wins, including one result from its own internal benchmark, so teams should test the comparison against their workloads before treating it as a ranking.